Hi all,

I'm releasing a small, coarse, splitting geocoder @ where 2.0 today (3:30
PM, "launch presentations", 4:15 office hours). I figured I'd preview it
here in case anyone on the list is at where and interested to hear me pitch
it in real life.

Here's the source: https://github.com/foursquare/twofishes

I'm going to post prebuilt jar binaries later so people can build the index
& serve without dealing with scala at all.

It's based off geonames, with flickr bounding boxes. There's a legacy v0
version written in python, and the current maintained version in scala. It
takes about an hour on a decently fast machine to build a full world
geonames index. At the end, you end up with ~3gb of hfiles which you can
use to server. A warm server can do 500qps @ 10ms latency.

I've been thinking about OSM data as well, but the political hierarchy
doesn't really seem to be present there. If someone wants to help me figure
out how to extract political feature relations from OSM, I'm interested. At
the very least, I could use OSM bounding boxes where present.

Foursquare is looking into putting some missing data into
geonames/twofishes, I think we might have better SE Asia coverage, though I
haven't had much time to look into it.

We'll be hosting this in our API at reasonable rate limits for people
getting started, as well as integrating it into all our API calls so that
existing calls that require a lat/lng centroid or boundingbox can use
near=QUERY instead. That's launching later this week, once Where is over
and I can be in front of a computer to watch the launch.

thanks,
--dave
_______________________________________________
Geowanking mailing list
[email protected]
http://geowanking.org/mailman/listinfo/geowanking_geowanking.org

Reply via email to